Flexible, logarithmic distribution, tag cloud for Jekyll.

jekyll.tag-cloud.rb

# Copyright (C) 2011 Anurag Priyam - MIT License

module Jekyll

  # Jekyll plugin to generate tag clouds.
  #
  # The plugin defines a `tag_cloud` tag that is rendered by Liquid into a tag
  # cloud:
  #
  #     <div class='cloud'>
  #         {% tag_cloud %}
  #     </div>
  #
  # The tag cloud itself is a collection of anchor tags, styled dynamically
  # with the `font-size` CSS property. The range of values, and unit to use for
  # `font-size` can be specified with a very simple syntax:
  #
  #     {% tag_cloud font-size: 16 - 28px %}
  #
  # The output is automatically formatted to use the same number of decimal
  # places as used in the argument:
  #
  #     {% tag_cloud font-size: 0.8  - 1.8em  %}  # => 1
  #     {% tag_cloud font-size: 0.80 - 1.80em %}  # => 2
  #
  # Tags that have been used less than a certain number of times can be
  # filtered out from the tag cloud with the optional `threshold` parameter:
  #
  #     {% tag_cloud threshold: 2%}
  #
  # Both the parameters can be easily clubbed together:
  #
  #     {% tag_cloud font-size: 50 - 150%, threshold: 2%}
  #
  # The plugin randomizes the order of tags every time the cloud is generated.
  class TagCloud < Liquid::Tag
    safe = true

    # tag cloud variables - these are setup in `initialize`
    attr_reader :size_min, :size_max, :precision, :unit, :threshold

    def initialize(name, params, tokens)
      # initialize default values
      @size_min, @size_max, @precision, @unit = 70, 170, 0, '%'
      @threshold                              = 1

      # process parameters
      @params = Hash[*params.split(/(?:: *)|(?:, *)/)]
      process_font_size(@params['font-size'])
      process_threshold(@params['threshold'])

      super
    end

    def render(context)
      # get an Array of [tag name, tag count] pairs
      count = context.registers[:site].tags.map do |name, posts|
        [name, posts.count] if posts.count >= threshold
      end

      # clear nils if any
      count.compact!

      # get the minimum, and maximum tag count
      min, max = count.map(&:last).minmax

      # map: [[tag name, tag count]] -> [[tag name, tag weight]]
      weight = count.map do |name, count|
        # logarithmic distribution
        weight = (Math.log(count) - Math.log(min))/(Math.log(max) - Math.log(min))
        [name, weight]
      end

      # shuffle the [tag name, tag weight] pairs
      weight.sort_by! { rand }

      # reduce the Array of [tag name, tag weight] pairs to HTML
      weight.reduce("") do |html, tag|
        name, weight = tag
        size = size_min + ((size_max - size_min) * weight).to_f
        size = sprintf("%.#{@precision}f", size)
        html << "<a style='font-size: #{size}#{unit}' href='/tags.html##{name}'>#{name}</a>\n"
      end
    end

    private

    def process_font_size(param)
      /(\d*\.{0,1}(\d*)) *- *(\d*\.{0,1}(\d*)) *(%|em|px)/.match(param) do |m|
        @size_min  = m[1].to_f
        @size_max  = m[3].to_f
        @precision = [m[2].size, m[4].size].max
        @unit      = m[5]
      end
    end

    def process_threshold(param)
      /\d*/.match(param) do |m|
        @threshold = m[0].to_i
      end
    end
  end
end

Liquid::Template.register_tag('tag_cloud', Jekyll::TagCloud)

I've replaced line 82 by

html << "<a style='font-size: #{size}#{unit}' href='" << Jekyll.configuration({})['baseurl'] << "/tag/#{slug}' title='#{count} topics'>#{name}</a>\n"

This gives absolutee links to tag pages. But it also shows Configuration file: /home/moose/Downloads/MartinThoma.github.io/_config.yml very often. I guess the tag cloud is generated every time a new page is generated. This is hopelessly inefficient.

How could we create the tag cloud once and only insert the generated HTML where needed?

Break this into two plugins: a generator plugin instead and a filter plugin (See the Jekyll documentation on writing plugins). The generator will run first and only once, construct a hash containing all the data you need to make the tag cloud (e.g name, weight pairs). Then you can define filter tag as you do above, but just have it read the hash from the generator rather than constructing the data.

For sorting the tags alphabetically, switching: weight.sort_by! { rand } for:
weight.sort_by! { |m| m.first.downcase } worked for me. But I really know nothing about ruby so...

What is i have to write in _config.yml file to start it? Just my second day with Jekyll so i full of love as well as chaos in fact))
P.S. Nice nickname :-)